= Cutdown Canonical Serialization Policy

== Goal

Different conforming implementations must be able to produce byte-identical serialized parse results from the same input and profile.

== Canonical Envelope

Canonical serialized output contains:

- ``cutdown_version``
- ``profile``
- ``ast``
- ``diagnostics``

== Ordering Rules

- Object keys are emitted in lexicographic order.
- Arrays preserve semantic order from parsing.
- ``diagnostics`` are sorted by (``start_line``, ``start_col``, ``code``).
- ``profile.extensions`` are sorted lexicographically.

== Normalization Rules

- No insignificant whitespace in canonical JSON.
- Strings use JSON standard escaping.
- Null fields are included only when required by schema; optional absent fields are omitted.
- Numeric fields use canonical decimal form (no leading plus, no unnecessary leading zeros).

== AST/Diagnostics Shape Stability

- Node field names and diagnostic field names are contract-bound.
- Renaming, adding required fields, or changing field semantics follows versioning policy.

== Conformance Requirements

Conformance corpus must include canonical serialized snapshots for representative cases, including malformed/recovery cases with diagnostics.

== Related

- [``../spec/017-canonical-form.md``](/latest/#s-17)
- [``./conformance-policy.md``](/latest/policies/conformance-policy)
- [``./diagnostics-policy.md``](/latest/policies/diagnostics-policy)